This paper describes the results of research on the application of a dynamically downscaled seasonal climate prediction system at Queensland Department of Natural Resources, Mines and Energy (NRME). The NRME seasonal climate prediction system consists of NCEP T40/L18 Global Climate Model and double nested CSIRO Regional Climate Model (RCM). The RCM at 75 km horizontal grid spacing and 18 vertical levels was integrated for years 1965-2003 for domain covering Australian region producing an ensemble of 15 downscaled integrations. Further downscaling using the double nesting approach at 15 km horizontal grid spacing over Queensland region has been also completed and produced regional predictions at high resolution. A study has been completed using this system forced with observed sea surface temperatures for period 1965-2003. The skill of the modelling system was evaluated in terms of spatial and temporal variability and the capacity to simulate extremes.

Major findings were: (a) use of Regional Climate Model increased the accuracy of simulated rainfall; and (b) when the downscaled output was linked to the spatial grazing simulation system for Australia (Aussie-Grass) a comparison with the current Southern Oscillation Index based operational systems revealed better performance than the benchmark statistical forecasting system when evaluated in terms of production and resource condition.

The NRME seasonal climate prediction system has been run operationally using predicted sea surface temperatures since September 1998 at monthly intervals. Ensemble predictions with lead-times of 7 months allow a probabilistic approach to risk management. The AGCM model output forms part of the IRI Net Assessment This forecasts are utilised worldwide and successfully provided an early warning of increased chance of drought in eastern Australia in 2002. The skill of dynamically downscaled seasonal forecasts using predicted sea surface temperatures has been evaluated and the results will be presented.